SEMINARS |
|
Spring 2007
STATISTICS
COLLOQUIUM
Friday, April 20, 2007
2:00-3:00—Talk,
Yost, Room 101
Junfeng Liu, PhD
Statistics Department
West Virgina University
A Self-Calibration Algorithm for Binary Model Goodness of Fit Test
Given probability function Pr(X,beta) along with a large population of continuous predictor X and response Y, there is an obvious motivation for testing accountability of model Y~Bernoulli(Pr(X,beta)). This paper develops a powerful profile based algorithm, classification-error-rate-self-calibration (CERSC), for global goodness of fit test. We consider classifying the subject with covariate X into Y=1 or 0 by comparing Pr(X,beta) with threshold tau, Pr(X,beta)>tau leads to Y=1 and Pr(X,beta)
|
| |
